Network Economist

Wherever you look on the web today, it’s hard not to notice the buzz over SOPA and PIPA.  Wikipedia, Reddit, Boingboing, and other sites are shut down in protest against these bills, which would create blacklists of websites accused of supporting piracy, and force network providers to prevent access to them.  Much like the notice-and-take-down procedures in the DMCA, this is a system ripe for abuse.  Small companies and individuals simply won’t have the resources to counter accusations of piracy and may be forced to shut down.  Over at Harvard Business Review, James Allworth and Maxwell Wessel argue persuasively that these bills are about entrenched big companies fighting off new innovators.  Even more so, this battle is about two different views of the internet: is it a democratic tool where anyone can express themselves, or is it a delivery mechanism for big content companies, who want to closely account for everything you do.  To learn more, the Electronic Frontier Foundation has collected a great deal of useful information about these bills.

What exactly is network economics?  As the world becomes more deeply networked, an increasing number of researchers fall into this space, drawing on both computer science and economic methods.  What makes this field unique, however, and why do we need people to study it?  To explain how the internet is special, I’d like to share a slide I made for a recent lecture.

If you’re a computer scientist, the internet is unique among engineered systems in one major way: it is federated, meaning that different actors own different parts of the system.  This means that behavior depends as much on human interests as it does on protocols.  Many technologies can’t even reach users unless network operators are motivated to deploy them in their network equipment.  This is famously difficult to do, a condition often referred to as the internet impasse.

If you view it from the perspective of traditional economics on the other hand, the internet brings a couple of new elements.  Some of these have been relatively easy to handle.  The internet can be congested by too much traffic.  Internet technologies also exhibit what economists call network effects – they become more valuable to each user as more people use them.  In these cases, economists have been successful at adapting existing models to the internet setting.  There is one property, however, that has proven much harder to deal with: the internet’s rich topology of interconnections.

Models of competition over network topologies are challenging to create, but necessary to gain a realistic picture of the internet.  Without them, we cannot see the fundamentally new ways in which networked systems can behave, and we risk transplanting conventional wisdom from traditional markets to the network setting.  Overcoming these limitations cannot be done with easy tweaks to existing models.  As we progress, network economics needs to continue developing its own modeling tradition, and this will make it appear increasingly like its own separate discipline.

Google has long been seen as a stout advocate for net neutrality. So eyebrows were raised earlier this week when the Wall Street Journal revealed the outlines of a deal that the internet giant has secretly pursued with major ISPs. According to documents reviewed by the journal, Google wishes to place servers directly within providers’ domains, giving geographically dispersed users quicker access to content. Content caching of this sort is already provided by companies like Akamai to significantly enhance the end-user experience. These systems, however, are freely available on non-discriminatory terms. Google’s, it appears, would favor the company’s own content and give it an advantage over competitors.

Has Google tripped over its own principles of net neutrality? Not so, according to the many commentators that rushed to the firm’s defense, including Lawrence Lessig, David Isenberg, Ed Felten, and Timothy Lee. Net neutrality, they say, is really about preventing ISPs from prioritizing some types of network traffic over others. Enhancing network applications through content caching is an entirely different matter.

From a technological standpoint, of course, content caching and traffic prioritization have enormous differences. But it is economic factors that drive network innovation and investment, and this is the basis by which we must judge Google’s proposals. Unfortunately, none of the company’s sympathizers has explained how the incentives behind content caching differ from those behind traffic prioritization. Indeed, the pitfalls of Google’s approach are strikingly familiar: an ISP controls an “upgrade” that enhances network service (either content caching, or the rights to attach a company’s own content caches). When agreements are secret, the ISP can tailor prices to each customer, punishing those that might threaten the provider’s profitable businesses. Even when a customer is not a direct competitor, an ISP will be tempted to assess the customer’s profitability and extract as much money as possible. Such a climate could have a chilling effect on innovation.

Read the rest of this entry »

Watching the debate over net neutrality can be a mystifying experience. Several years in, and despite a mounting canon of scholarship, experts continue to disagree on what neutrality means, exactly. As a result, arguments that seem to oppose each other are often incomparable, simply addressing entirely different concepts. To make any sense of such arguments, one must first ask, what kind of neutrality? In fact, there are three prominent strands within today’s literature, each of which formulates neutrality along a different dimension:

1. Are multiple service tiers allowed? Service tiers, or classes, are a way to give different treatment to different network traffic. While some computer scientists maintain that ever-increasing bandwidth will render all service classes obsolete, many others believe that tiers could play a useful role into the future, providing tailored connections to applications with different requirements. Service classes could be used to provide steady throughput for online video, high security for sensitive information, or high dependability for mission-critical applications like tele-surgery.

2. How much discrimination is allowed in each service tier? This was the central issue in the recent FCC action against Comcast. ISPs can discriminate by charging different prices to different users, or, as in Comcast’s case, by keeping prices equal but offering different service qualities to different users or applications.

3. Who pays whom? Traditionally, money flows from end-users at the edge of the Internet towards the center, handed from ISP to ISP. Large providers would like to bypass this structure to contract directly with end-users. This is the least understood dimension of net neutrality, even though changes here would have major ramifications.

Read the rest of this entry »

Last week, the Federal Communications Commission (FCC) weighed in on the longstanding dispute between broadband provider Comcast, and users that share files over the BitTorrent application. In an order that has yet to be released, the Commission found that Comcast had violated its users’ rights by blocking BitTorrent traffic and misleading the public about its actions. BitTorrent users claimed a victory as Comcast was ordered to end its discriminatory practices. Unfortunately, as we discussed in our previous post, the precedent that emerges from the order is a rather narrow one. To put it briefly, discriminatory practices that don’t further reasonable network management goals are forbidden. Clever broadband providers can probably work around this restriction, though, by using intelligent congestion control or prioritization techniques, which can similarly degrade file sharing applications. Although the FCC has left itself some space to redefine “reasonable” network management in the future, its stance appears timid when compared with other strategies it could have followed. Here, we present three alternate frameworks the Commission could have relied on.

1. Commissioner Michael Copps argues that the FCC should adopt an explicit principle of non-discrimination. Discrimination is a tricky concept to define, and sketching the boundary of such a principle would admittedly be a challenge. Nevertheless, a non-discrimination principle would guard against many potential abuses that are likely to follow in the wake of Comcast. For example, such a rule could be used to pursue an ISP that reserved a lot of bandwidth for normal web traffic, leaving little for file sharers. Comments by FCC chairman Kevin Martin suggest that this type of prioritization is permitted under the current order.

2. The FCC could alternately resurrect the old Title II framework for broadband providers, which existed before 2002. This framework recognized broadband service as a “pure” and “transparent” transmission of data, and imposed a number of common carrier rules, including an obligation not to discriminate.

3. Both the above approaches, as well as the FCC’s, miss out on one crucial development: telecommunications convergence. As network technology has progressed, the old divisions between internet, television, and telephone have steadily disappeared. All these services now use the same basic technology, and what little divisions remain are largely artificial. A forward-looking approach for the FCC must therefore recognize that there is essentially just one network.

This kind of framework would illuminate the root cause behind Comcast’s discrimination: a company with a lot of power in one market (for broadband access), trying to leverage that power into a second market (for network content). Comcast could be regulated much like any other firm in this situation: by requiring non-discriminatory access for all network services. In other words, television content and internet content should have equal opportunities to purchase network transport. This would eliminate today’s perverse incentives to limit internet bandwidth, lest YouTube becomes a better substitute for television. It would also invite companies to innovate in television service and invent new hybrid models of entertainment.

Such a framework remains farfetched today, and telecommunications firms everywhere are working hard to maintain the illusion that there are separate networks. Nevertheless, the long-term trend of convergence is unmistakable, and we may one day see our laws catch up to it.

The Federal Communications Commission (FCC) has long favored deregulation and industry consensus over active enforcement on the Internet. As broadband provider Comcast discovered last week, however, the right provocation can still cause this watchdog to bare its teeth. The stimulus in this case was Comcast’s interference with the popular BitTorrent file sharing application. Embracing the appearance of guilt, Comcast first denied ever meddling with its customers’ communications, until tests by the Associated Press (AP) and the Electronic Frontier Foundation (EFF) showed that BitTorrent traffic was indeed being blocked. Comcast then argued that its system only kicked in during periods of network congestion, until this claim was similarly disproved. Even after such antics, the FCC was nearly split, with two out of five commissioners defending the ISP.

The FCC’s action is good news for the internet’s file sharers. Comcast must now end its discriminatory practices and report back to the Commission to explain what new practices will replace them. The FCC also upbraided Comcast over its dishonesty. After months of getting the runaround, file sharers can be forgiven for taking some delight in this. But how much will this order protect file sharers in the future?

Pessimists will note that Comcast’s brazenness left the FCC with little choice but to act against it. The ISP defended its blocking policy as necessary for network management, but the FCC rejected this explanation, pointing out that Comcast’s system was ill-suited for this purpose. For example, blocking took place regardless of the level of congestion. This evidence left Comcast’s apologists sounding shrill. One dissenter worried that the commissioners, lacking engineering degrees, were not equipped to judge how networks were managed (no doubt implying that the FCC should leave the Internet alone entirely). Another dissenter fretted about the chilling effect the order would have on ISPs trying to tackle child pornography and other illegal content (even though it was precisely the lack of such a productive purpose that landed Comcast in trouble).

Read the rest of this entry »

In slides and on blackboards everywhere, network researchers are fond of depicting the internet as a cloud – and the analogy is no accident. The internet protocols were designed to run on top of a variety of underlying substrates and cover up their differences. The result is a simple interface that underlies the internet’s success story, but – as the cloud metaphor emphasizes – conceals its internal structure. This has been the nature of the network since its birth, but an important shift is currently under way. If emerging trends continue, the idea of a network cloud may soon seem antiquated.

Internet users gained an early bit of extra visibility into the network last October, when The Associated Press (AP) reported that Comcast was interfering with the popular BitTorrent file-sharing application. By comparing the data packets sent by one network user to those received at the other end, the AP discovered that the sequences didn’t always match up as expected. The recipient would sometimes receive a special “reset” packet, which appeared to be sent by the other computer, but was instead inserted by Comcast itself. By “forging” such packets, Comcast would effectively disrupt the communications of its own customers.

These revelations added fuel to the ongoing net neutrality debate. Comcast defended its tactics, saying they were unavoidable to keep BitTorrent traffic from overwhelming the network. Critics noted that discriminating against file sharers conveniently restrains a nascent competitive threat to Comcast’s profitable television business. The neutrality question will not be resolved anytime soon. Whatever the ultimate outcome, the debate’s greatest legacy may arise by spawning technologies that make the network more transparent.
Read the rest of this entry »